Clinical experience with immunomodulatory therapies in patients with sepsis has been disappointing to date. This patient group may not be uniform however, and important factors that differ between patients may influence the fundamental effects of such therapies. We have previously shown that prophylactic therapy with G-CSF increases endotoxin clearance and improves outcome with intraperitoneal sepsis in a canine model. We have also found, however, that in contrast to the beneficial effects of rG-CSF pretreatment in our canine model, similar pretreatment in rats administered intrabronchial bacteria is harmful. The nature of these different effects is unclear. If species is not an issue, other differences in our canine and rat models such as site and severity of infection and dose of G-CSF are several possibilities. Such factors might also be at work in clinical studies evaluating immunomodulatory therapies. In order to study these factors, we have now developed an experimental system using a small animal model and a multifactorial design, to evaluate the effects of site and severity of infection, and dose of immunomodulatory therapy. With this model, we are studying the differential effects of placebo, low or high dose G-CSF pretreatment during mild or severe bacterial intravenous, intraperitoneal or intrabronchial infection. Early results suggest that site and severity of infection may fundamentally change tide effects of G-CSF therapy, being beneficial in some cases but harmful in others.